Back To Course Home

Make Your First GAN Using PyTorch

0% completed

✨Introduction

AI is Exploding!Who Is This Course For?

✨PyTorch Basics

Why PyTorch?Pytorch Tensors Automatic Gradients with PyTorch Computation Graphs 📚Learning Points in PyTorch Basics

🖥️ Project: Computation Graph and Gradient Calculation

🍀 Challenge: Setting Up a Computation Graph Solution Review: Setting Up a Computation Graph 🍀 Challenge: Finding Gradient Solution Review: Finding Gradient

✨First PyTorch Neural Network

MNIST Image Dataset Simple Neural Network Visualizing Training The MNIST Dataset Class Training Our Classifier Querying Our Neural Network Simple Classifier Performance 📚Learning Points in First PyTorch Neural Network

✨Refinements

Activation Function

Optimisation Function

Combined Refinements

📚Learning Points in Refinements

🖥️ Project: Fashion MNIST Classification Model

The Fashion MNIST Dataset Class 🍀 Challenge: Define the Neural Network Architecture Solution Review: Define the Neural Network Architecture The Fashion MNIST Classifier Performance

✨CUDA Basics

NumPy vs. Python Nvidia CUDA Using CUDA in Python 📚Learning Points in CUDA Basics

Course Assessment

Make Your First GAN: Part 1 - Exam

✨The GAN Idea

Generating Images Adversarial Training Training a GAN 📚Learning Points in The GAN Idea

✨Simple 1010 Pattern

Generating Real Data Building The Discriminator Training and Testing The Discriminator Building The Generator Training The 1010 Pattern GAN 📚Learning Points in Simple 1010 Pattern

🖥️ Project: A Simple 01110 Pattern GAN

Generating Real Data 🍀 Challenge: Building the Discriminator Solution Review: Building the Discriminator 🍀 Challenge: Building the Generator Solution Review: Building the Generator Training the 01110 Pattern GAN

✨Handwritten Digits

The MNIST Dataset Class The MNIST Discriminator The MNIST Generator Training the MNIST GAN Mode Collapse Improving GAN Training Experimenting with Seeds 📚Learning Points in Handwritten Digits

🖥️ Project: Fashion MNIST GAN

🍀 Challenge: Building the Discriminator Solution Review: Building the Discriminator 🍀 Challenge: Building the Generator Solution Review: Building the Generator Training The Fashion MNIST GAN

✨Human Faces

The Colour Images CelebA Dataset The Dataset Class The Discriminator The Generator Training The Human Faces GAN 📚Learning Points in Human Faces

Course Assessment

Make Your First GAN: Part 2 - Exam

✨Convolutional GANs

Introduction to Convolutional GANs MNIST CNN CelebA CNN Experimentation 📚Learning Points in Convolutional GANs

🖥️ Project: A Fashion MNIST CNN

🍀 Challenge: Define the Convolution Neural Network Architecture Solution Review: Define the Convolution Neural Network The Fashion MNIST CNN Classifier Performance

✨Conditional GANs

Introduction to Conditional GANs The Discriminator The Generator Training the MNIST cGAN Conditional GAN Results 📚Learning Points in Conditional GANs

🖥️ Project: Fashion MNIST Conditional GAN

🍀 Challenge: Define the Discriminator Solution Review: Define the Discriminator 🍀 Challenge: Define the Generator Solution Review: Define the Generator Training the Fashion MNIST cGAN

Course Assessment

Make Your First GAN: Part 3 - Exam

✨Conclusion

🏆 A Quick Overview 🚀 What's Next?

✨Appendix

Appendix 0: Author's Notes Appendix 1: Google Colab Workspace Appendix A: Ideal Loss Values Appendix B: GANs Learn Likelihood Appendix C: Convolution Examples Appendix D: Unstable Learning Appendix E: Acknowledgments

Combined Refinements

Let's combine all the refinements we learned in the previous lessons.

We'll cover the following

Be wise in combining refinements!

Let’s combine all these refinements, BCE loss, leaky ReLU activation, Adam optimiser, and layer normalization.

Be wise in combining refinements!

Because the BCE loss can’t accept values outside the range 0 to 1, which can emerge from a leaky ReLU, we can have a sigmoid after the last layer, but keep a LeakyRELU after the hidden layer.

Get hands-on with 1200+ tech skills courses.